Meta May Limit AI Token Spending Per Engineer

Meta could soon introduce spending limits on how much AI computing power each engineer can use, according to Instagram chief Adam Mosseri. As artificial intelligence becomes a bigger part of software development, the cost of running AI models is rising quickly, prompting companies to think more carefully about how these resources are used.

Speaking on Lenny’s Podcast, Mosseri said he can imagine a future—possibly within the next year or two—where an engineer’s AI token usage could cost the company as much as that employee’s salary. If that happens, Meta may need to place caps on AI token spending for individual employees.

AI tokens are the units used to process prompts and responses when interacting with large language models. As more developers rely on AI-powered coding assistants and other tools, token usage has become a major expense for technology companies.

According to Mosseri, AI token costs should eventually be treated like any other business resource. Companies already manage budgets for payroll, cloud infrastructure, hardware, and operating expenses, and AI usage may soon require the same level of oversight.

He explained that businesses constantly make decisions about how to allocate limited resources such as GPUs, CPUs, storage, RAM, and labeling budgets across different teams. In the future, token budgets could become another category that managers need to distribute strategically.

Mosseri believes any future spending limits would not be applied equally across all employees. Instead, the amount of AI tokens available to each engineer would likely depend on the company’s confidence that the individual can use those resources efficiently and generate a positive return on investment.

For now, Meta has not introduced AI token caps for its workforce. However, the company has already taken steps to reduce unnecessary AI spending after discovering that some internal projects were consuming large amounts of tokens without delivering meaningful value.

One example was an internal AI token spending leaderboard, which encouraged employees to compete based on token usage. Meta eventually shut down the leaderboard after AI costs reportedly put the company on track to spend billions of dollars on AI token processing in 2026.

The issue isn’t unique to Meta. Several major technology companies have also been forced to rethink how they use AI because of rapidly increasing operating costs.

Uber recently faced its own AI budgeting challenge after exhausting its planned AI coding budget for 2026 by April. Meanwhile, Microsoft reportedly reduced expenses by canceling Claude Code licenses and encouraging engineers to use its own Copilot CLI development tool instead.

Despite today’s rising costs, Mosseri expects the situation to improve over time. He believes competition among AI model providers will eventually trigger pricing wars, making AI services more affordable as companies compete to attract developers and enterprise customers.

Until then, Meta is focusing on cutting waste rather than limiting innovation. Mosseri noted that the company has already eliminated what he described as “silly things” that consumed large amounts of AI tokens without producing significant business value.

As AI becomes an everyday tool for software engineers, companies are entering a new phase where managing AI resources could become just as important as managing salaries, infrastructure, or cloud computing budgets. Meta’s approach suggests that while AI adoption will continue to grow, businesses will increasingly expect developers to use these powerful—and expensive—tools responsibly.

Also read : General Fusion Shares Jump 40% After Nasdaq Debut